Skip to content

omamobile-device-husky 0.1.15: Wi-Fi power save on by default - #836

Merged
derekross merged 1 commit into
derekross:mainfrom
Somnius:husky-wifi-powersave-1115
Nov 18, 2026
Merged

derekross merged 1 commit into
derekross:mainfrom
Somnius:husky-wifi-powersave-1115

Conversation

@Somnius

@Somnius Somnius commented Nov 16, 2026

Copy link
Copy Markdown
Contributor

The 8 Pro's Wi-Fi runs with power save off: Omarchy's /etc/NetworkManager/conf.d/omarchy-wifi-powersave.conf sets wifi.powersave=2, and iw says "Power save: off". This turns it on for husky (wifi.powersave=3, as on the 6 Pro).

Wi-Fi power save on the 8 Pro

Left: battery current with the cable out and the screen off, 2026-11-14, six steps of about 4 minutes with power save off, off, on, off, on, off (16 readings each). Step medians 537, 541, 532 and 543 mA with it off, 509 and 508 mA with it on, so about 30 mA less with it on. Right: 200 pings at 0.2 s from the desktop to the phone, off, on, off again: average 2.26, 2.52 and 2.27 ms, maximum 5.8, 71.2 and 6.1 ms, no loss in any. So power save adds an occasional latency spike, the usual trade.

Why the file is /etc/NetworkManager/conf.d/zz-omamobile-wifi-powersave.conf and not in /usr/lib: NetworkManager reads /usr/lib first and /etc after, so Omarchy's /etc file wins over anything in /usr/lib. I checked on the phone with NetworkManager --print-config: a file in /etc named 90-omamobile-... or omamobile-... still read 2, zz-omamobile-... in /etc read 3, and the same zz file in /usr/lib read 2 again. It's in backup= so a changed copy survives an upgrade.

Tested: the package's test/all (26 ok) and tools/pr-check. On the phone by hand, the same file in /etc made --print-config read 3, iw reported power save on, and it stayed on after a reconnect. Not tested: the installed package itself, hours of use, hotspot, and Wi-Fi after resume with power save on.

The version says 0.1.14 since main is at 0.1.13; #806 also uses 0.1.14, so please set whatever fits at merge.

@derekross

Copy link
Copy Markdown
Owner

Thank you, Lef. The measurement is the right kind (alternating steps, medians, and the latency cost shown next to the saving), and the reasoning about the file's place is correct: I rebuilt your three cases with scratch config directories and NetworkManager 1.58, and got 2, 2 and 3 exactly as you describe. Omarchy's file sets wifi.powersave = 2 on purpose, for latency and for Intel firmware that drops the link, so reversing it for one battery-powered phone inside its device package is the right scope.

Before I merge it:

  1. Version and order. omamobile-device-husky 0.1.14: one sleep state, so a refused sleep isn't tried twice (#619) #806 is merging as 0.1.14, the number your phone has from that branch. Please rebase this onto main after it and make it 0.1.15 (the version file and the title). The only conflict is the README: your new row sits directly above the sleep row omamobile-device-husky 0.1.14: one sleep state, so a refused sleep isn't tried twice (#619) #806 rewrote; keep both.

  2. Three README wordings.

    • "A file of the same setting in /etc/NetworkManager/conf.d wins over this one": only a file whose name sorts after zz-omamobile-wifi-powersave.conf does, as your own 90-omamobile-… test showed. The simple way for an owner to turn it off is to edit this file, which backup= keeps; please say that instead.
    • "The one setting it ships is Wi-Fi power save, in the table above." is in the Wi-Fi section twice.
    • "as on the Pixel 6 Pro" (README and the file's comment): the 6 Pro's package asks for the same value from /usr/lib, which, by what you found, loses to Omarchy's file. "the value the Pixel 6 Pro's package asks for" is safe. I'll check the 6 Pro side separately; thank you for finding that.
  3. One run of the installed package, with one sleep. You said clearly that the package itself and resume are untested, and I'd like both before this becomes the default on all three 8 Pros, since your omamobile-device-husky 0.1.14: one sleep state, so a refused sleep isn't tried twice (#619) #806 sleep had power save off. With the package built from the rebased branch:

    • pacman -Q omamobile-device-husky, pacman -Qo /etc/NetworkManager/conf.d/zz-omamobile-wifi-powersave.conf, NetworkManager --print-config | grep -A2 '^\[connection' and iw dev wlp1s0 get power_save after a reconnect;
    • one attended systemctl suspend on the charger with Wi-Fi connected, as you did for omamobile-device-husky 0.1.14: one sleep state, so a refused sleep isn't tried twice (#619) #806: suspend_stats before and after, the entry and exit lines, whether Wi-Fi is back and pinging, and anything new from dmesg | grep -iE 'brcmf|pcie';
    • tools/phone-smoke from that boot.

A line in husky's NOTES with the two sets of numbers (30 mA, and the 71 ms spike) would be welcome too, so the trade stays on record.

The Pixel 8 Pro's Wi-Fi read "Power save: off", and the phone drew
about 30 mA more on battery that way than with it on (step medians of
532 to 543 mA off over four steps, 508 and 509 mA on over two,
2026-11-14).

The package now ships wifi.powersave=3, the value the Pixel 6 Pro's package asks for,
as /etc/NetworkManager/conf.d/zz-omamobile-wifi-powersave.conf,
and lists it in backup= so an upgrade keeps a copy the user has changed.
NetworkManager applies it when a connection comes up.

Tested: the package's test/all on the desktop, and by hand on the phone (the package itself is not installed there): the same file in /etc/NetworkManager/conf.d, named zz-omamobile-wifi-powersave.conf with wifi.powersave=3, made NetworkManager --print-config read 3 where it read 2 before, iw reported power save on, and it stayed on after a reconnect. 200 pings at 0.2 s to the phone over Wi-Fi with power save off, on and off again: average 2.26, 2.52 and 2.27 ms, maximum 5.8, 71.2 and 6.1 ms, no loss in any. So it adds a latency spike here. 30 mA less on battery was measured earlier (2026-11-14, two 4-minute steps with it on between four with it off). Not tested: the installed package, hours of use, hotspot, resume from sleep.

The file is installed in /etc and named zz-omamobile-wifi-powersave.conf on purpose: omarchy-settings ships /etc/NetworkManager/conf.d/omarchy-wifi-powersave.conf with wifi.powersave=2. Checked on the phone with NetworkManager --print-config: a file in /etc named 90-omamobile-... or omamobile-... still read 2, a zz-omamobile-... file in /etc read 3, and the same zz file in /usr/lib read 2 again, because NetworkManager reads /usr/lib first and /etc after it.
@Somnius
Somnius force-pushed the husky-wifi-powersave-1115 branch from c81289b to d4ce18c Compare November 18, 2026 06:54
@Somnius Somnius changed the title omamobile-device-husky 0.1.14: Wi-Fi power save on by default omamobile-device-husky 0.1.15: Wi-Fi power save on by default Nov 18, 2026
@Somnius

Somnius commented Nov 18, 2026

Copy link
Copy Markdown
Contributor Author

Thanks @derekross. Rebased onto main after #806 and it's 0.1.15 now (version file and title), README rows kept both.

The three wordings:

  • the override sentence now says only a file that sorts after zz-omamobile-wifi-powersave.conf wins, and that the simple way to turn it off is to edit this file, which backup= keeps;
  • the duplicate sentence is gone;
  • "as on the Pixel 6 Pro" is now "the value the Pixel 6 Pro's package asks for", in the README, the file's comment and the commit message.

husky's NOTES has a Power save line with both sets of numbers (about 30 mA less, and the 71.2 ms ping max).

The installed-package run with one attended suspend isn't done yet, my phone is off right now. I'll post the pacman -Q, pacman -Qo, --print-config, iw, the suspend stats and dmesg lines, and a tools/phone-smoke report here when I've run it.

@Somnius

Somnius commented Nov 18, 2026

Copy link
Copy Markdown
Contributor Author

@derekross the run of the installed package, on my 8 Pro, on the charger, Wi-Fi connected.

Package built from this branch's head (d4ce18c) and installed with pacman -U:

omamobile-device-husky 0.1.15-1
/etc/NetworkManager/conf.d/zz-omamobile-wifi-powersave.conf is owned by omamobile-device-husky 0.1.15-1
[connection]
wifi.powersave=3

One thing I didn't expect: right after the install a reconnect still gave Power save: off. NetworkManager reads conf.d only at start or on a reload, and --print-config reads the files, not the running daemon. After nmcli general reload conf and a reconnect: Power save: on. A reload in the install script would make it apply at once. The husky install script has to match cheetah's byte for byte (the package test checks it), so I left it out of this PR. Happy to add it to every device package's script in its own PR if you want that.

The suspend, with a 120 s RTC alarm:

  • suspend_stats success/fail: 0/0 before, 1/0 after
  • PM: suspend entry (s2idle) at 10:32:12, PM: suspend exit at 10:34:12
  • Wi-Fi associated again at 10:34:21, NetworkManager activated at 10:34:22, then pings to the router went through
  • new brcmf|pcie lines after the wake: brcmf_p2p_create_p2pdev: timeout occurred, add iface p2p-dev-wlp1s0 type 10 failed: err=-5 (the same one Colin is watching for H2), and one brcmf_join_v1: ... failed: firmware error -14 before the join that worked, plus a few Received unexpected monitor pkt

To be straight about what else was on the phone: my local sleep blocker for charging (it holds a block inhibitor while the charger is online) was stopped for this run and started again after. /etc/systemd/sleep.conf.d/ is empty. The #765 watchdog drop-in (/etc/systemd/system/service.d/10-test-765.conf, WatchdogSec=0) is still there, I'll remove it once those runs are done.

tools/phone-smoke --here from that boot:

Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-push1-gdc329ba6088b, slot b, up 4930s
storage guard (required) ok 62 partitions; only userdata and slot b's boot images writable
root (required) ok ext4 on userdata, 200G free
kernel health (required) ok no oops or BUG; 1 WARN, 32 error lines in dmesg (29 known, of them UFS dme-set 0x4021 (s2idle) x1, 3 new)
new dmesg errors not ok join (chanspec, bytes) failed: firmware error x3
systemd units ok none failed
USB gadget (required) ok bound to 11210000.usb
display ok card2-DSI-1 connected, 1344x2992
GPU ok render node on panthor
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock
touch ok SALIXP Touchscreen
buttons ok gpio-keys
Wi-Fi ok wlp1s0 on brcmfmac, link up
Bluetooth ok hci0, powered true
battery ok 87% Charging, 39 °C
sound cards ok googlehuskyaocs
thermal ok hottest zone 56 °C

Packages: hyprland 0.57.0-3.2, mesa 1:26.2.4-1.4, omamobile-device-husky 0.1.15-1, omamobile-shell 0.2.107-1, omarchy 4.0.6-1, quickshell 0.3.2.r7.g2b4501a-1. The kernel is my own local build (omamobile plus my charger test commits), not the published one.

@derekross

Copy link
Copy Markdown
Owner

Thank you, Lef. Version, rebase and the three wordings are all as asked, the NOTES line has both sets of numbers, and the run of the installed package is what I needed: the package owns the file, --print-config reads 3, iw reads on, and Wi-Fi came back 9 seconds after a two-minute sleep with nothing new in the log beyond the four lines we already know. Merging it.

Your finding that the setting only applies after NetworkManager reloads its configuration is worth one sentence in the README: an updated phone gets it from its next boot. You were right to leave a reload out of the pacman script. Could you add that sentence in your next husky PR, and take out "Not yet booted from the package" there once you have rebooted on it?

One thing for the record, since the smoke report shows it: this ran on your local kernel, in the hours you describe on #761. I've written about that on linux-omarchy-mobile#266; nothing to do here.

@derekross
derekross merged commit 9c24fa2 into derekross:main Nov 18, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants